Problems URIs Don’t Solve * * but people think they should
description
Transcript of Problems URIs Don’t Solve * * but people think they should
Problems URIs Don’t Solve**
** but people think they shouldLarry MasinterXerox CorporationPalo Alto Research Center
http://www.parc.xerox.com/masinterhttp://purl.org/NET/masinter
[email protected], [email protected], [email protected], [email protected], [email protected], [email protected], ...
•2
Guide to the talk
• update WWW4 talk:“problems URIs don’t solve”
• a few more modern examples• a few new problems• new directions for solutions
•3
Problems URIs don’t solve
• Dangling URLs• Vanity, commerce, & ownership• Using name for the unnamable
•4
Recap: The UR- storyUR-here, UR-there, UR-UR-Everywhere
• URL: Uniform Resource Locator– where’s the resource?– ubiquitous
• URN: Uniform Resource Name– location independent name
• URC: Uniform Resource Characteristics– title, author, information
• URI: Uniform Resource Identifier– any of the above
•5
The Dangling URL problem
• The Web4Lib Problem“I created a list of interesting sites for my
users… six months later, half the links didn’t work! What can I do?”
“Don’t worry, the URI committee is solving this problem… Location independent names!”
• The URN solution: location-independent names:– Create a URN for the resource– Links include URNs instead of URLs– Tada! Problem solved!
•6
Reasons URNs didn’t solve the problem
• Stuff moves• Stuff goes away• Over time, everything goes away• Authority and trust issues
•7
Stuff Moves (in Mysterious Ways)
• Organizations split– AT&T => bellcore, belllabs, lucent, att– How to find old research reports? Who maintains the old namespace?
• Organizations rename– Resource naming based on authority/organizational name– The old organization name is someone else’s trademark– The current “owner” of the name isn’t an authority for the old name
• Organizations merge– names and authority change– The new organization now controls the – URN for “NAI” (Network Associates) vs. “McAfee” for virus files
• Grad students eventually graduate– http://www.ics.uci.edu/pub/ietf/http/ home page– University authority or credentials don’t migrate– Employees leave the company– Organizational hierarchical names are transient
•8
Stuff goes away
• The student isn’t interested• The company goes out of business• The country disappears/merges/splits• Eventually, almost everything disappears• URN requirement: “500 years”
– What namespaces remain from 1499?– How do we resolve them?
•9
Internet Scale Naming must scale over Time
• Almost all of the approaches to scaling don’t deal with time scale
• A simplistic solution: remember URL + timestamp
• “The document that was at http://www.ics.uci.edu/pub/ietf/http/ in August of 1999.”
• Date URLs by date in referring material• No help for resolution
•10
Vanity, Commerce, and Ownershipnow a well-recognized problem
• Now a well-recognized problem• Generic terms: pimples.com• Shared trademarks
– apple the fruit, apple music, apple computer– delta the airline, delta the faucet-maker
• Migration of ownership– MTV.com
• Approaches: WIPO, Authority & policy, new TLDs, relative names
•11
Most approaches don’t solvethe problems
• Need another level of indirection?Hacking DNS TLDs doesn’t help
• Solve only this problem,but don’t solve the other naming problems
•12
Naming the unnamable
• The “tv:” scheme– Tune your TV to a particular channel
• What’s a channel? Is Tv:5 a name? A uniform name?• Name a station?• “name” a segment of time?
• The “data:” scheme– Every level of indirection needs a way around it– Include the data instead of a pointer to the data– data:image/gif;base64,R0lGODdhMAAwAPAAAAAAAP///
ywAAAAAMAAwAAAC8IyPqcvt3wCcDkiLc7C0qwyGHhSWpjQu5yqmCYsapyuvUUlvONmOZtfzgFz ByTB10QgxOR0TqBQejhRNzOfkVJ+5YiUqrXF5Y5lKh/DeuNcP5yLWGsEbtLiOSp a/TPg7JpJHxyendzWTBfX0cxOnKPjgBzi4diinWGdkF8kjdfnycQZXZeYGejmJl ZeGl9i2icVqaNVailT6F5iJ90m6mvuTS4OK05M0vDk0Q4XUtwvKOzrcd3iq9uisF81M1OIcR7lEewwcLp7tuNNkM3uNna3F2JQFo97Vriy/Xl4/f1cf5VWzXyym7PHhhx4dbgYKAAA7
– data:application/vnd.webtv.tune,channel=upn,date=8/16/1999,time:23:00
•13
Are Fragments names?
• The stuff after the # in your URI• Is it part of the URI or not?• Part of name depends on media being
pointed into• Redirection, indirection to/from components• Naming interacts with document model
•14
Names for non-English
• Other Western European languages: accents– http://www.métro.paris.fr/métro
• Hebrew and Arabic– Direction marks– vowel markers
• Asian languages– Private characters for names– Variations with accents for Thai, Urdu
• Even for English: capitalization creating trade names from generic terms
•15
CNRP: Common Name ResolutionHow to use “common names” for resources?
• draft-popp-cnrp-goals-00.txt– define a protocol
• for browsers & search sites to use• for ‘name space’ providers to offer• simpler than ‘search’• more featureful than ‘string lookup’
– Implies a service• reviews nature of services, revenue & motivation of
participants• Can’t be solved independently of business & social
issues
– Please review & participate